EDPB Defines AI Data Use Rules as UN Calls for Global AI Safeguards

The European Data Protection Board (EDPB) this week responded to a request from Ireland’s Data Protection Authority, outlining a framework for assessing when personal data may be legitimately used to train AI models without breaching GDPR.

The opinion reaffirms that AI models relying on personal data must meet strict anonymity thresholds, ensuring individuals cannot be identified with anything more than an “insignificant” likelihood.

The EDPB introduced a three-stage test to determine legitimate interest: identifying the interest, assessing its necessity, and ensuring it does not override individuals’ fundamental rights.

The Board emphasized transparency and accountability in AI development, while also giving national data protection authorities the discretion to assess compliance on a case-by-case basis.

"Models developed with data processed illegally may not be deployed," the opinion stated.

Mixed Reactions from Stakeholders

Tech industry representatives, such as the Computer & Communications Industry Association (CCIA), welcomed the opinion, saying it enables AI models to access quality data for accuracy and bias mitigation.

However, digital rights advocates, including Itxaso Dominguez de Olazabal of EDRi, expressed concerns.
They questioned the practicality of the anonymity thresholds and warned that broad discretion for national regulators could lead to inconsistent enforcement across the EU.

“Harmonisation is key to ensuring digital rights are upheld universally,” Dominguez de Olazabal said.

The EDPB also announced forthcoming guidelines on web scraping—automated data extraction for AI model training—an issue critical as AI development increasingly depends on vast datasets.

Global Urgency for AI Governance

Meanwhile, UN Secretary-General António Guterres urged the Security Council on Thursday to take decisive action in establishing international AI safeguards.

Addressing the risks of unregulated AI, Guterres stressed that delays could threaten global peace, particularly in the military domain.

“Every moment of delay increases the risk for us all,” Guterres warned, highlighting concerns over AI’s integration with military applications and nuclear systems.
“No country should design or use AI in armed conflict that violates humanitarian and human rights laws,” he added.

Guterres noted AI’s dual nature, citing its role in predicting climate-related displacements and detecting landmines while warning of its dangers in autonomous decision-making, surveillance, and disinformation.

The UN chief emphasized that AI should not erode trust or worsen inequalities, calling for collaborative international governance.

“Unprecedented global challenges call for unprecedented global cooperation,” Guterres said, urging a ban on lethal autonomous weapons and tighter controls on AI’s militarization.

Guterres concluded by encouraging the Security Council to lead in building a “safe, secure, and inclusive AI future.”